Denoise

This category focuses on the foundational plumbing for AI agents. Projects are either creating generalized infrastructure like `HKUDS/CLI-Anything` for tool use, or specialized, Claude-centric orchestration platforms like `ruvnet/ruflo`.

• tinyhumansai/openhuman

  An ambitious project aiming to create a personal, private AI. While the description is vague, its focus on privacy and local control taps into growing concerns about centralized AI systems. It's one to watch for architectural patterns.

  rising⭐ 20k
• HKUDS/CLI-Anything

  This project enables AI agents to use existing command-line interface (CLI) tools. This is a significant piece of infrastructure, as it potentially makes decades of robust, battle-tested software accessible to agents without needing new APIs.

  new⭐ 37k
• TauricResearch/TradingAgents

  A framework for building multi-agent systems for financial trading. It's a niche but complex application of LLM agents, demonstrating their use in specialized domains requiring collaboration and data analysis.

  cooling⭐ 77k
• ruvnet/ruflo

  An agent orchestration platform specifically for Anthropic's Claude. Its existence and popularity are a strong signal of the market consolidating around Claude as a preferred backend for multi-agent systems.

  cooling⭐ 53k

This category is overwhelmingly dominated by the Claude ecosystem. We see a split between serious attempts to create production-grade agent methodologies (`humanlayer/12-factor-agents`) and a larger volume of low-effort, personal prompt collections.

• mattpocock/skills

  A personal collection of Claude prompts from a well-regarded developer. Its popularity highlights a trend where developers curate and share their own 'skill files' to standardize and improve agent behavior for their specific workflows.

  repeated⭐ 94k
• Imbad0202/academic-research-skills

  A set of prompts tailored for academic research tasks. This demonstrates the specialization of agent skills for specific professional domains, moving beyond generic coding or writing assistance into more structured, expert workflows.

  rising⭐ 14k
• multica-ai/andrej-karpathy-skills

  A single prompt file that encapsulates Andrej Karpathy's advice on LLM coding. This is an interesting form of knowledge distillation, attempting to bake expert intuition and best practices directly into an agent's core instructions.

  repeated⭐ 137k
• obra/superpowers

  This project presents a more structured, methodological approach to agent skills, rather than just a list of prompts. It represents an effort to professionalize agent development and make it more systematic and repeatable.

  rising⭐ 198k
• anthropics/claude-plugins-official

  The official plugin directory from Anthropic. This is a major ecosystem development, signaling Anthropic's move to control and standardize its platform's extension model, akin to the launch of an app store. It legitimizes the plugin ecosystem.

  new⭐ 20k
• anthropics/financial-services

  An empty repository from Anthropic. Its trending status is likely due to speculation or name recognition. It currently provides no value but indicates areas Anthropic may be exploring for enterprise solutions.

  repeated⭐ 26k
• msitarzewski/agency-agents

  A collection of prompts framed as a complete 'AI agency'. This appears to be a low-effort project riding the hype cycle around autonomous agents, packaging existing ideas without contributing new technology.

  new⭐ 101k
• yikart/AiToEarn

  A repository with a title suggesting AI-based monetization schemes. Lacking specific technical details, it likely falls into the category of low-quality content or prompt collections aiming to attract attention during a hype cycle.

  new⭐ 15k
• addyosmani/agent-skills

  From another well-known developer, this repository focuses on 'production-grade' skills for coding agents. The emphasis on production readiness is key, showing a push towards building reliable, maintainable agent-based systems.

  cooling⭐ 44k
• humanlayer/12-factor-agents

  An attempt to define a set of principles for building robust, production-ready LLM applications, analogous to the classic '12-Factor App' methodology. It's a sign of the field maturing and seeking established engineering discipline.

  new⭐ 21k

These projects address a core limitation of LLMs: statefulness and context size. `rohitg00/agentmemory` provides persistent memory for long-running tasks, while `colbymchenry/codegraph` offers a token-efficient way to inject codebase context.

• rohitg00/agentmemory

  Provides persistent, long-term memory for AI agents. This is a critical infrastructure component for building stateful agents that can learn and maintain context across multiple sessions, moving beyond simple request-response interactions.

  rising⭐ 14k
• colbymchenry/codegraph

  A tool that creates a local knowledge graph of a codebase for AI agents. This is a practical solution to the context window limitation, aiming to reduce token usage and improve the agent's understanding of large, complex projects.

  rising⭐ 6.0k

A mix of general-purpose tools and AI-specific utilities. `CloakHQ/CloakBrowser` is critical for agent web automation, while `rtk-ai/rtk` and `millionco/react-doctor` address the cost and quality problems of AI-generated code. Major players like Meta (`pyrefly`) and Microsoft are also present.

• CloakHQ/CloakBrowser

  A modified version of Chromium designed to evade bot detection systems. This is a critical enabling technology for web automation agents, as their utility depends on their ability to access websites without being blocked.

  rising⭐ 16k
• oven-sh/bun

  A fast, all-in-one JavaScript runtime and toolkit. Its continued presence on the trending list reflects its significant adoption and impact on the JavaScript ecosystem, consolidating various development tools into a single, high-performance binary.

  new⭐ 92k
• rtk-ai/rtk

  A command-line proxy written in Rust that aims to dramatically reduce LLM token consumption for common development tasks. Like other similar tools today, it directly addresses the significant operational cost of using large models.

  new⭐ 51k
• facebook/pyrefly

  A new, fast type checker for Python from Meta. As a new entrant in the crowded Python tooling space from a major tech company, it's worth monitoring to see if it gains traction against established tools like MyPy and Pyright.

  new⭐ 6.3k
• Diolinux/PhotoGIMP

  A patch for the open-source image editor GIMP to make its interface and shortcuts resemble Adobe Photoshop. This is a long-standing community effort to lower the barrier to entry for users familiar with industry-standard proprietary software.

  new⭐ 11k
• millionco/react-doctor

  A tool specifically designed to identify and fix common errors in React code generated by AI agents. This addresses a critical problem: the poor quality and non-idiomatic nature of much AI-generated code, acting as a linting layer for agent output.

  new⭐ 10k
• Fincept-Corporation/FinceptTerminal

  A desktop-style financial analysis application. It represents a category of complex, data-intensive applications being built with modern web technologies, moving beyond simple websites or mobile apps.

  cooling⭐ 22k
• pascalorg/editor

  A tool for creating and sharing 3D architectural projects. Its presence indicates continued interest in browser-based tools for specialized, creative domains that have traditionally been dominated by desktop software.

  new⭐ 16k
• apernet/hysteria

  A proxy and censorship circumvention tool. Its popularity often correlates with geopolitical events or changes in internet filtering policies in various regions, reflecting an ongoing demand for unrestricted internet access.

  new⭐ 21k
• frappe/erpnext

  A popular open-source ERP system. Trending for this kind of mature project often indicates a new major release, a significant migration of users from proprietary systems, or growing interest in self-hosted business software.

  new⭐ 34k
• microsoft/ai-agents-for-beginners

  A curriculum for learning how to build AI agents, provided by Microsoft. This is a strong indicator of the mainstreaming of agent technology, with major platform vendors now investing in educational materials to onboard developers.

  new⭐ 64k

Access and cost are the primary drivers here. Both trending repos are wrappers or proxies (`Alishahryar1/free-claude-code`, `decolua/9router`) designed to provide free access to proprietary models like Claude, indicating significant developer demand constrained by pricing.

• Alishahryar1/free-claude-code

  A tool providing free access to the Claude Code model via various interfaces. Its persistent popularity underscores the high demand for powerful coding models and suggests that official API costs are a significant barrier for many developers.

  repeated⭐ 26k
• decolua/9router

  A proxy that routes AI requests to various free model providers, with features like auto-fallback and token reduction. It's an arbitrage tool designed to minimize costs, highlighting how central token economy is to developers building on LLMs.

  cooling⭐ 12k

This category shows a dual focus on automated video generation (`AIDC-AI/Pixelle-Video`, `HKUDS/ViMax`) and novel, non-visual sensing methods like `ruvnet/RuView`'s use of WiFi signals. It's a move towards both automating creative output and expanding sensory input for AI.

• AIDC-AI/Pixelle-Video

  An automated short video generation engine. This project is part of a larger trend of using AI to automate content creation pipelines, particularly for social media formats where quantity and speed are important.

  cooling⭐ 18k
• ruvnet/RuView

  This project uses commodity WiFi signals for spatial awareness and vital sign monitoring, without cameras. It's a novel application of existing technology for privacy-preserving presence detection, with potential uses in smart homes and healthcare.

  new⭐ 61k
• Anil-matcha/Open-Generative-AI

  A self-hostable, open-source suite for AI image and video generation that explicitly states it has no content filters. This project appeals to users seeking unrestricted creative freedom and control over their generative AI tools.

  cooling⭐ 16k
• heygen-com/hyperframes

  A library for rendering video from HTML, designed for programmatic use by AI agents. This provides a structured, declarative way for agents to create video content, moving beyond simple text or image generation.

  cooling⭐ 20k
• HKUDS/ViMax

  An 'agentic' video generation framework that assigns roles like 'director' and 'screenwriter' to different agents. It's an experiment in structuring the creative process for generative AI to achieve more coherent and complex outputs.

  new⭐ 5.2k

The focus is on open-source intelligence (OSINT). While `Z4nzu/hackingtool` is a generic toolkit, `soxoj/maigret` and `BigBodyCobain/Shadowbroker` represent more specialized tools for collecting data on individuals and high-value assets, respectively.

• Z4nzu/hackingtool

  An aggregation of various hacking scripts and tools. While popular, these 'all-in-one' toolkits often have low signal, attracting beginners. They are more of a curated list than a novel piece of technology.

  cooling⭐ 76k
• soxoj/maigret

  An open-source intelligence (OSINT) tool for finding a person's online presence across thousands of websites from a single username. It's a specialized and effective tool for digital investigation and reconnaissance.

  cooling⭐ 29k
• BigBodyCobain/Shadowbroker

  An ambitious open-source intelligence (OSINT) platform for tracking high-value assets like private jets and satellites. It represents a trend towards democratizing access to sophisticated global monitoring and intelligence-gathering capabilities.

  new⭐ 8.1k